A computer simulation analysis of the accuracy of partial genome sequencing and restriction fragment analysis in estimating genetic relationships: an application to papillomavirus DNA sequences
© Qiao and Weigel; licensee BioMed Central Ltd. 2004
Received: 12 March 2004
Accepted: 27 July 2004
Published: 27 July 2004
Determination of genetic relatedness among microorganisms provides information necessary for making inferences regarding phylogeny. However, there is little information available on how well the genetic relationships inferred from different genotyping methods agree with true genetic relationships. In this report, two genotyping methods – restriction fragment analysis (RFA) and partial genome DNA sequencing – were each compared to complete DNA sequencing as the definitive standard for classification.
Using the Genbank database, 16 different types or subtypes of papillomavirus were selected as study samples, because numerous complete genome sequences were available. RFA was achieved by computer-simulated digestion. The genetic similarity of samples, based on RFA, was determined from the proportion of fragments that matched in size. DNA sequences of four specific genes (E1, E6, E7, and L1), representing partial genome sequencing, were also selected for comparison to complete genome sequencing. Laboratory error was not taken into account. Evaluation of the correlation between genetic similarity matrices (Mantel's r) and comparisons of the structure of the derived dendrograms (partition metric) indicated that partial genome sequencing (for single genes) had higher agreement with complete genome sequencing, achieving a maximum Mantel's r = 0.97 and a minimum partition metric = 10. RFA had lower agreement, with a maximum Mantel's r = 0.60 and a minimum partition metric = 18.
This simulation indicated that for smaller genomes, such as papillomavirus, partial genome sequencing is superior to restriction fragment analysis in representing genetic relatedness among isolates. The generalizability of these results to larger genomes, as well as the impact of laboratory error, remains to be demonstrated.
Precise estimation of genetic relatedness between isolates of a microorganism is important for determination of phylogenetic relationships, which has important applications in studies of disease transmission [1, 2]. The definitive standard for assessing genetic relatedness among organisms is the complete genome sequence of nucleotide bases . However, nucleotide sequencing is expensive and time-consuming, thus, generally it is impractical for use in most investigations, particularly when a large number of samples is analyzed.
Currently, one genotyping technique used frequently as an alternative to complete genome sequencing is restriction fragment analysis (RFA), in which restriction endonuclease enzymes cleave the genome at specific sites, producing DNA fragments that are then separated by size using electrophoresis . The percentage of fragments matching in size has been commonly used as an index to represent the genetic similarity between samples [5, 6]. The accuracy of RFA in determining the true genetic relationships can be influenced by several factors, including the number of restriction enzymes used, the specific enzymes selected for DNA digestion, and laboratory conditions [7–9].
Another common alternative to complete genome sequencing is partial genome sequencing, i.e., the nucleotide sequencing of a particular gene or segment of the genome [8, 10]. The gene or genome segment is often targeted by polymerase chain reaction (PCR). Selection of an appropriate gene or region for analysis is critical for accurately representing phylogenetic relationships [11, 12].
In a comparison of RFA and partial genome sequencing with respect to their similarity in interpreting a disease outbreak caused by pseudorbabies virus in a swine producing region in Illinois, USA, both genotyping methods generated similar conclusions about patterns of spread of the virus . However, the accuracy of each genotyping method in representing the complete genome was not evaluated.
Restriction fragment analysis detects genetic variation by surveying specific endonuclease restriction sites over the entire genome; in contrast, partial genome sequencing detects genetic variation by comparing nucleotide bases from a specific region of the genome. Each method detects a different dimension of genetic variation, and each can detect only a proportion of the genetic variation present in the entire genome. Therefore, it is important to determine which method, using partial information, provides a more accurate estimation of genetic relatedness.
The primary purpose of this study was to compare both restriction fragment analysis and partial genome sequencing to complete genome sequencing, with regard to their agreement in estimating genetic relationships and in reconstructing phylogenies under the ideal conditions of absence of laboratory error. Computer simulation of the genotyping analysis was conducted, using completely sequenced papillomavirus isolates obtained from Genbank.
Sequence lengths and fragment size distribution for papillomavirus samples obtained from Genbank
Type of Papillomavirus
Genbank Accession Number
Length of Complete Genome (bps)
Length of Genes (bps)
Fragment Size Distribution (digested by MaeI enzyme1)
Number of Fragments
Median Fragment Size (bps)
5% Percentile (bps)
95% Percentile (bps)
Similarity of restriction fragment analysis to complete genome sequencing in estimating genetic relatedness between papillomavirus samples
Number of Enzymes
Similarity of partial genome sequencing to complete genome sequencing in estimating genetic relatedness between papillomavirus samples
Mantel's Correlation Coefficient
Sequencing entire genomes is impractical in most investigations of genetic relationships. The computer simulation conducted here determined that compared to restriction fragment analysis, partial genome sequencing had higher agreement with complete genome sequencing in estimating genetic relatedness and greater similarity in the topology of the dendrograms of phylogenetic relationships derived from these estimates. These results using papallomavirus sequences with a genome length averaging less than 8 kb, indicate that for microorganisms with small genomes, partial genome sequencing targeting genes comprising approximately 20–25% of the total genome length can provide a very good estimate of genetic relatedness. The topological structure of phylogenetic trees was also stable for partial genome sequencing, particularly for the most closely related samples. The degree to which these results generalize to larger genomes is unknown, in part because microorganisms with large genomes are rarely, if ever, sequenced in their entirety. There are also other considerations in selecting partial genome sequencing as a genotyping method, such as presence of the gene in all isolates, and sufficient variability to differentiate isolates . In addition, whether genetic variation is random or due to natural selection needs to be taken into account , because in the latter case genetic dissimilarity may not reflect time since divergence, thus making it more difficult to infer evolutionary relationships, which are important for making inferences about pathogen transmission. These limitations should be considered as well for restriction fragment analysis.
One might expect that increasing genome size would diminish the advantage of partial genome sequencing compared to restriction fragment analysis. As total genome size increases, the number of restriction sites cut by restriction enzymes is expected to increase, providing more fragments and more genetic information for estimating genetic relatedness at no increased cost. This also needs to be taken into account in the selection of a genotyping method. However, it has been argued that if a gene is selectively neutral (i.e., variations are not subject to natural selection), it is only the length of the gene sequenced, not the ratio of sequenced gene length to genome size, that is important for determining the degree of divergence from a common ancestor . To the extent that these conditions are satisfied, the results of this study indicate that specific gene sequencing is likely to provide a better estimate of genetic relationships than restriction fragment analysis of the complete genome under a wider variety of genome sizes.
The general conditions under which partial genome sequencing is more accurate than restriction fragment analysis in representing true genetic relatedness have not been addressed in the analysis conducted here. However, another study from our laboratory , using simulated genomes of various size with different nucleotide substitution rates, and varying degrees of genetic diversity among samples, found that only under conditions of both short partial genome sequence length and low rates of nucleotide substitution did RFA provide a more accurate topological reconstruction of phylogenetic relationships than did partial genome sequencing; the degree of genetic diversity among samples did not affect the advantage partial genome sequencing had in accurately depicting phylogenetic relationships. Thus, whether one is investigating the genetic relatedness among samples collected from a single disease outbreak or a diverse collection of samples from different times and geographic regions, under most conditions partial genome sequencing will represent genetic relationships more accurately than does RFA. Genotyping using partial genome sequencing and phylogenetic reconstruction (using the neighbor-joining algorithm) have become standard for several virus species, including not only papillomavirus [16, 17], but also human immunodeficiency virus , classical swine fever virus , porcine reproductive and respiratory syndrome virus , and foot-and-mouth disease virus .
The simulated genotyping conducted here assumed no error of measurement. The sources of error in restriction fragment analysis are well known [22–24]. Fragments of similar size in the same lane of a gel may be indistinguishable, thus appearing to form one fragment. Fragments of small size may be undetectable. The relationship between migration distances and fragment size may be affected by variation in gel density both between and within gels. There are also differences in measurement error between laboratories [25, 26]. These deficiencies are accounted for by use of marker DNA fragments of known nucleotide base pair length to assist in estimating cleaved DNA fragment sizes; however, acknowledgement of remaining error of measurement of the size of detectable fragments is inherent in the application of a tolerance range for considering fragments of similar but different sizes as a "match" . Laboratory error is also inherent in partial genome sequencing . With the commonly used polymerase chain reaction (PCR) methodology for detection and amplification of genes for sequencing, there can be error in primer development because primer sites may not be specific to the gene sequences or too specific to demarcate all occurrences of the gene. Heterogeneity of amplified DNA, due to replication error, recombination, low primer specificity, or impurity of the template can result in a failure to produce consistent sequencing results. In the comparison of the degree of similarity of DNA sequences between samples, alignment of sequences with unequal sequence lengths due to deletion or duplication, or the management of inverted sequences presents additional challenges for estimating genetic similarity and phylogenetic affinity . The relative magnitude of sources of error in RFA versus partial genome sequencing is unknown and, thus, the conclusions presented here are those based upon the assumption of the absence or minimization of laboratory error.
In practical terms, laboratory error and cost need to be taken into account in the selection of a genotyping method. However, when the impact of these factors is minimized, the computer simulation analysis conducted here indicates that partial genome sequence becomes the preferred alternative for representing genetic relationships.
For small genomes, partial genome sequencing of target genes comprising 20–25% of the total genome provides a more accurate estimate of genetic relatedness and more accurate representation of evolutionary and transmission histories than does restriction fragment analysis and thus is indicated to be the preferred genotyping method for phylogenetic reconstruction under these conditions. The degree to which these results are generalizable to larger genomes and conditions of laboratory error remains to be determined.
Sample DNA sequences
The source of information on nucleotide sequences was the Genbank database . The organism selected for analysis was papallomavirus, for which a moderately large number of isolates with complete genome sequences was available. Human, bovine, canine, and chimpanzee papillomaviruses were considered. Among human papillomavirus (HPV) with complete genome sequencing available, 12 samples were selected at random: HPV 4, 6a, 6b, 20, 24, 49, 63, 13, 29, 32, 54, and 26. For bovine papillomavirus (BPV), complete genome sequences were available for BPV1, BPV2, and BPV4. Because the E1 gene of BPV1 (of interest for partial genome sequencing) could not be located, only BPV2 and BPV4 were chosen and included in the study. One type of canine oral papillomavirus (caninePV) and one type of common chimpanzee papillomavirus (chimPV) were available in the database, and these were chosen. Thus, a total of 16 types or subtypes of papillomaviruses that have been completely sequenced and stored in Genbank were used (Table 1).
The complete DNA sequences of the 16 papillomavirus samples were aligned using ClustalW software . The genetic distances among these sequences were then calculated using the Kimura correction [31, 32].
Computer simulated restriction fragment analysis
Restriction endonuclease enzymes
Commonly used restriction endonuclease enzymes were selected , based on the following criteria: (1) Only enzymes with 4-base pair recognition sites were selected, in order to produce a sufficient number of fragments for analysis. (2) Among enzymes having the same recognition site, only one was selected. (3) For simplicity, enzymes with multiple recognition sites were excluded. Using these criteria, 15 restriction enzymes were included (AccII, AciI, AluI, BsuRI, CviRI, HapII, HhaI, MaeI, MaeI, MboI, MseI, NlaIII, RsaI, TaqI, TspEI).
Simulated digestion of each papillomavirus DNA sample by each restriction enzyme was conducted using the DIGEST program . The resulting restriction fragments for each sample were sorted by size (number of nucleotide base pairs).
Calculation of genetic distances
Based on the distribution of restriction fragment sizes, the genetic similarity between any two papillomavirus samples was calculated for each restriction enzyme using the Dice coefficient [5, 6]: Sxy = 2nxy/(nx+ny), where nxy is the number of fragments matching in size for samples x and y, and nx and ny are the number of fragments in samples x and y, respectively. Then, Dxy = 1-Sxy was calculated as a distance measure. Pairwise distances between samples were computed for each individual enzyme. Also, pairwise distances were obtained for up to 4 enzymes, by using for each condition (2, 3, and 4 enzymes) the fragment size distributions for 30 randomly selected combinations of enzymes, and calculating the composite distance .
Partial genome sequence analysis
The E1, E6, E7, and L1 genes, which have been of interest in studies of papillomavirus, were used for estimating genetic relatedness. The ClustalW program  was used for sequence alignment, and the genetic distances (with the Kimura correction) were calculated for each gene.
Agreement between genotyping methods
Correlation between distance matrices
The matrix of genetic distances based on complete DNA sequences was considered the definitive standard. The genetic distance matrices based on RFA and partial genome sequencing were compared to complete genome sequencing by calculating Mantel's coefficient of correlation between matrices (Mantel's r) .
Comparison of phylogenetic trees
The genetic distance matrices for RFA, partial genome sequencing, and complete genome sequencing were used to construct phylogenetic trees, using the Neighboring-joining algorithm , as implemented by MEGA software . Trees were rooted at the midpoint between the most distantly related samples . Bootstrap values indicating stability of tree topology were added to trees based on partial and complete genome sequencing . The trees based on RFA and specific gene sequences were compared to the tree for complete genome sequencing, by using the COMPONENT software  to calculate the partition metric, which measures the difference in tree topology [41, 42]. A lower value of partition metric indicates greater topological similarity.
BQ designed the investigation, collected the data, conducted the data analysis, and wrote the manuscript. RW identified the problem to be investigated, provided statistical guidance, assisted in interpretation of results, and edited the final drafts of the manuscript. Both authors read and approved the final manuscript.
List of abbreviations
- Mantel's r:
Mantel's coefficient of correlation between matrices
polymerase chain reaction
restriction fragment analysis
- Tenover FC, Arbeit RD, Goering RV, Mickelsen PA, Murray BE, Persing DH, Swaminathan B: Interpreting chromosomal DNA restriction patterns produced by pulse-field gel electrophoresis: criteria for bacterial strain typing. J Clin Microbiol 1995, 33: 2233–2239.PubMed CentralPubMedGoogle Scholar
- Salamon H, Behr MA, Rhee JT, Small PM: Genetic distances for the study of infectious disease epidemiology. Am J Epidemiol 2000, 151: 324–334.View ArticlePubMedGoogle Scholar
- Upholt WB: Estimation of DNA sequence divergence from comparison of restriction endonuclease digests. Nucleic Acids Res 1977, 4: 1257–1265.PubMed CentralView ArticlePubMedGoogle Scholar
- Dowling TE, Moritz C, Palmer JD, Riesenberg LH: Nucleic acids III: analysis of fragments and restriction sites. In Molecular Systematics (Edited by: Hillis DM, Moritz C, Mable BK). Sunderland, Massachusetts: Sinauer 1996, 249–320.Google Scholar
- Lynch M: The similarity index and DNA fingerprinting. Mol Biol Evol 1990, 7: 478–484.PubMedGoogle Scholar
- Call DR, Hallett JG, Mech SG, Evans M: Considerations for measuring genetic variation and population structure with multilocus fingerprinting. Mol Ecol 1998, 7: 1337–1346. 10.1046/j.1365-294x.1998.00454.xView ArticleGoogle Scholar
- Römling U, Grothues D, Heuer T, Tümmler B: Physical genome analysis of bacteria. Electrophoresis 1992, 13: 626–631.View ArticlePubMedGoogle Scholar
- Holt RJ, Strike P, Bruce DK: Phylogenetic analysis of tnpR genes in mercury resistant soil bacteria: the relationship between DNA sequencing and RFLP typing approaches. FEMS Microbiol Lett 1996, 144: 95–102. 10.1016/0378-1097(96)00345-XView ArticlePubMedGoogle Scholar
- Arens M: Methods for subtyping and molecular comparison of human viral genomes. Clin Microbiol Rev 1999, 12: 612–626.PubMed CentralPubMedGoogle Scholar
- Takewaki S, Okuzumi K, Manabe I, Tanimura M, Miyamura K, Nakahara K, Yazaki Y, Ohkubo A, Nagai R: Nucleotide sequence comparison of the mycobacterial dnaJ gene and PCR-restriction fragment length polymorphism analysis for identification of mycobacterial species. Int J Syst Bacteriol 1994, 44: 159–166.View ArticlePubMedGoogle Scholar
- Russo CAM, Takezaki N, Nei M: Efficiencies of different genes and different tree-building methods in recovering a known vertebrate phylogeny. Mol Biol Evol 1996, 13: 525–536.View ArticlePubMedGoogle Scholar
- Olive DM, Bean P: Principles and applications of methods for DNA-based typing of microbial organisms. J Clin Microbiol 1999, 37: 1661–1669.PubMed CentralPubMedGoogle Scholar
- Goldberg TL, Weigel RM, Hahn EC, Scherba G: Comparative utility of restriction fragment length polymorphism analysis and gene sequencing to the molecular epidemiological investigation of a viral outbreak. Epidemiol Infect 2001, 126: 415–424. 10.1017/S0950268801005489PubMed CentralView ArticlePubMedGoogle Scholar
- Nei M, Kumar S: Molecular Evolution and Phylogenetics New York: Oxford University Press 2000.Google Scholar
- Qiao B: Investigation of the accuracy of partial genome sequencing and restriction fragment analysis in determination of genetic relationships: a computer simulation study. Ph.D dissertation, University of Illinois at Urbana-Champaign 2003.Google Scholar
- Chan S-Y, Bernard H-U, Ratterree M, Birkebak TA, Faras AJ, Ostrow RS: Genomic diversity and evolution of papillomaviruses in rhesus monkeys. J Virol 1997, 71: 4938–4943.PubMed CentralPubMedGoogle Scholar
- De Villiers E-M, Fauquet C, Broker TR, Bernard H-U, zur Hausen H: Classification of papillomaviruses. Virology 2004, 324: 17–27. 10.1016/j.virol.2004.03.033View ArticlePubMedGoogle Scholar
- Gao F, Yue L, Robertson DL, Hill SC, Hui H, Biggar RJ, Neequaye AE, Whelan TM, Ho DD, Shaw GM, Sharp PM, Hahn BH: Genetic diversity of human immunodeficiency virus type 2: evidence of distinct sequence subtypes with differences in virus biology. J Virol 1994, 68: 7433–7447.PubMed CentralPubMedGoogle Scholar
- Greiser-Wilke I, Fritzmeier J, Koenen F, Vanderhallen H, Rutili D, de Mia G-M, Romero L, Rosell R, Sanchez-Vizcaino JM, San Gabriel A: Molecular epidemiology of a large classical swine fever epidemic in the European Union in 1997–1998. Vet Microbiol 2000, 77: 17–27. 10.1016/S0378-1135(00)00253-4View ArticlePubMedGoogle Scholar
- Forsberg R, Oleksiewicz MB, Krabbe Petersen A-M, Hein J, Bøtner A, Storgaard T: A molecular clock dates the common ancestor of European-type porcine reproductive and respiratory syndrome virus at more than 10 years before the emergence of disease. Virology 2001, 289: 174–179. 10.1006/viro.2001.1102View ArticlePubMedGoogle Scholar
- Knowles NJ, Samuel AR: Molecular epidemiology of foot-and-mouth disease virus. Virus Res 2003, 91: 65–80. 10.1016/S0168-1702(02)00260-5View ArticlePubMedGoogle Scholar
- Goering RV, Duensing TD: Rapid field inversion gel electrophoresis in combination with a rRNA gene probe in the epidemiological evaluation of Staphylococci . J Clin Microbiol 1990, 28: 426–429.PubMed CentralPubMedGoogle Scholar
- Maslow JN, Mulligan ME, Arbeit RD: Molecular epidemiology: application of contemporary techniques to the typing of microorganisms. Clin Infect Dis 1993, 17: 153–164.View ArticlePubMedGoogle Scholar
- Tenover FC, Arbeit RD, Goering RV, the Molecular Typing Working Group of the Society for Healthcare Epidemiology of America: How to select and interpret molecular strain typing methods for epidemiological studies of bacterial infections: a review for healthcare epidemiologists. Infect Control 1997, 18: 426–439.Google Scholar
- Laber TL, Iverson JT, Liberty JA, Giese SA: The evaluation and implementation of match criteria for forensic analysis of DNA. J Forensic Sci 1995, 40: 1058–1064.View ArticlePubMedGoogle Scholar
- Duewer DL, Lalonde SA, Aubin RA, Fourney RM, Reeder DJ: Interlaboratory comparison of autoradiographic DNA profiling measurements: precision and concordance. J Forensic Sci 1998, 43: 465–471.View ArticlePubMedGoogle Scholar
- Gill P, Evett IW, Woodroffe S, Lygo JE, Millican E, Webster M: Databases, quality control and interpretation of DNA profiling in the Home Office Forensic Science Service. Electrophoresis 1991, 12: 204–209.View ArticlePubMedGoogle Scholar
- Hillis DM, Mable BK, Larson A, Davis SK, Zimmer EA: Nucleic acids IV: sequencing and cloning. In Molecular Systematics (Edited by: Hillis DM, Moritz C, Mable BK). Sunderland, Massachusetts: Sinauer 1996, 321–381.Google Scholar
- Genbank Database[http://www.ncbi.nlm.nih.gov/Genbank]
- Thompson JD, Higgins DG, Gibson TJ: CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res 1994, 22: 4673–4680.PubMed CentralView ArticlePubMedGoogle Scholar
- Kimura M: A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences. J Mol Evol 1980, 16: 111–120.View ArticlePubMedGoogle Scholar
- Kimura M: Estimation of evolutionary distances between homologous nucleotide sequences. Proc Natl Acad Sci USA 1981, 78: 454–458.PubMed CentralView ArticlePubMedGoogle Scholar
- Sambrook J, Fritsch EF, Maniatis T: Molecular Cloning – A Laboratory Manual 2 Edition New York: Cold Spring Harbor Laboratory Press 1989.Google Scholar
- Nakisa RC: DIGEST, version 1.0. London: Imperial College of Science, Technology and Medicine 1993. [http://iubio.bio.indiana.edu/soft/molbio/ibmpc]Google Scholar
- Nei M: Molecular Evolutionary Genetics New York: Columbia University Press 1987.Google Scholar
- Mantel N: The detection of disease clustering and a generalized regression approach. Cancer Res 1967, 27: 209–220.PubMedGoogle Scholar
- Saitou N, Nei M: The Neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol Biol Evol 1987, 4: 406–425.PubMedGoogle Scholar
- Kumar S, Tamura K, Jakobsen IB, Nei M: MEGA: Molecular Evolutionary Genetics Analysis Software, version 2.1. Tempe: Arizona State University 2001. [http://www.megasoftware.net]Google Scholar
- Swafford DL, Olsen GJ, Waddell , Hillis DM: Phylogenetic Inference. In Molecular Systematics (Edited by: Hillis DM, Moritz C, Mable BK). Sunderland, Massachusetts: Sinauer 1996, 407–514.Google Scholar
- Page RDM: COMPONENT, version 2.0. London: The Natural History Museum 1993. [http://taxonomy.zoology.gla.ac.uk/rod/cpw.html]Google Scholar
- Robinson DF, Foulds LR: Comparison of phylogenetic trees. Math Biosci 1981, 53: 131–147. 10.1016/0025-5564(81)90043-2View ArticleGoogle Scholar
- Penny D, Hendy MD: The use of tree comparison metrics. Syst Zool 1985, 34: 75–82.View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.